import-skilllisted

# Inputs Expected from the user: - **repo owner** (required), e.g. `mattpocock` - **repo name** (required), e.g. `skills` - **skill name** (optional) — the skill's directory name, e.g. `grilling`. Without it, list the repo's skills and ask which to import. Intent is either **link** (add a new entry) or **update** (re-pin an existing entry to a newer sha). Infer it from whether the skill already appears in `.claude-plugin/marketplace.json`; ask only if the user's words contradict what you find. # High-level behavior 1. Read `.claude-plugin/marketplace.json` and check whether the skill already has an entry. This decides link vs update. 2. Resolve the repo's default branch and its HEAD sha (see [Discovery](#discovery)). 3. Locate the skill's `SKILL.md` at that sha. Every candidate path must be confirmed to exist at the resolved sha — never a path from memory or from an existing entry. 4. Read the skill's frontmatter `name` and `description` at that sha. 5. Resolve the author's display name. 6. Write the entry (see [Entry shape](#entry-shape)) into the `plugins` array, then verify the file still parses as JSON. 7. Report: name, path, sha (short), and — on update — the old sha it moved from. Done when the entry exists with a full 40-character sha, its `path` is confirmed present at that sha, and the JSON parses. # Discovery `?` needs quoting in zsh — quote every URL that carries a query string. ```sh # Default branch + HEAD sha (the pin) gh api repos/<owner>/<repo> --jq '